Post on 31-May-2020
1
.Imaging Radar - .The One Sensor To Rule Them All.
.Imaging Radar - .The One Sensor To Rule Them All. Amit Benjamin, Texas Instruments
Agenda
Radar in Automotive Today - What is Missing?
Camera Vs Lidar Vs Radar - Strength & Weakness
Which Sensors in Which AV Level?
Imaging Sensor - Why is it a Game Changer?
Imaging Radar Use Case Analysis - Step Function in Radar performance
Radar in Automotive
Determine object range
Determine object (relative) velocity
Determine object azimuth & elevation angle
Object discrimination (multi-target capability)
Object classification (shape, size)
Robust against environmental conditions such as bad light & weather
3
1
2
R1
R2
Vrel_1
Vrel_2
az
el
Sounds Simple – But Something doesn’t work Right
Sounds Simple – But Something is missing
LiDAR vs Camera vs RADAR
LiDAR vs Camera vs RADAR
LiDAR vs Camera vs RADAR
LiDAR vs Camera vs RADAR
LiDAR vs Camera vs RADAR Industry is
working on size
reduction and
cost down
ACC
Rear View Camera & Surround View
& Blindspot
& Stereo Front Camera & LIDAR Level 5 system!
Does all Sensors are needed for level 2 & 3 ? For 4 & 5?
Level 2 & 3? Level 4 & 5?
Does all Sensors are needed for level 2 & 3 ? For 4 & 5?
TI Confidential – NDA Restrictions
Radar &
Camera
sensors
Does all Sensors are needed for level 2 & 3 ? For 4 & 5?
Why Imaging Radar is a Game Changer?
Imaging Radar – First RF CMOS Cascade solution
• Pedestrian detection at > 140m • Car detection at > 350m • Azimuth / Elevation* angular resolution < 1°
The four devices are synchronized and work as a single unit, coherently processing data from all the antennas
* Might require Super Resolution
Why Cascading?
BeamForming
Multiple TX antennas transmit simultaneously & coherently to create a focused beam.
Phase shifts across TX antennas can steer the beam in a desired direction
Coherent gain across the NTX antennas improves SNR (20log10(NTX) vs. 10log10(NTX) in MIMO)
Maximum detection range - 350m Vs 120m with single chip
16
𝑆𝑁𝑅𝑇𝑋_𝐵𝐹 ∝ 𝑁𝑇𝑋2
MiMo - Angle resolution (θres) is the minimum angle separation for two objects to appear as separate peaks in the angle-FFT.
Improving angle resolution
A 2 TX / 4 RX can achieve a theoretical angle resolution of only 15o
TI imaging Radar with 192 virtual channels (four chip cascade) can achieve down to 0.6 degree of Azimuth and Elevation* Angle resolution
17
𝜃res =2
𝑁𝑇𝑋𝑁𝑅𝑋
Why Cascading?
* Might required Super Resolution
“…radar system is programmed to ignore high-mounted objects such as road signs and, possibly, the flanks of a semi truck, to avoid undesired braking events…”
Why Imaging Radar is a Game Changer? Overhead bridges/tunnels
el
static truck stalled under a bridge
el = 1.4⁰ = arctan(1m/40m)
40m
1m
Today: Typical 5⁰* Angle Resolution Front Radar
Tan(5⁰ ) = X/40 X = 3.5meter
Imaging Radar < 1⁰ both Azimuth & Elevation
The Radar Sensor will become the primary sensor in the car
* Might required Super Resolution
Summary - Why Imaging Radar is a Game Changer?
1m
With Imaging Radar we are able to:
Improve the Angle Resolution by a factor
of ~ 𝑁𝑇𝑋𝑁𝑅𝑋
Improves SNR (20log10(NTX) vs.
10log10(NTX) in MIMO) to significantly
increase the range
Significantly improve AV level 2 & 3 system
cost by potentially Eliminate few of the sensor
Summary - Why Imaging Radar is a Game Changer?
1m
With Imaging Radar we are able to:
Improve the Angle Resolution by a factor
of ~ 𝑁𝑇𝑋𝑁𝑅𝑋
Improves SNR (20log10(NTX) vs.
10log10(NTX) in MIMO) to significantly
increase the range
Significantly improve AV level 2 & 3 system
cost by potentially Eliminate few of the sensor
Questions? Thank You